We compared the effects of flumazenil, bretazenil, and midazolam on schedule-controlled behavior in squirrel monkeys (N=4) responding under a multiple schedule conflict procedure. Midazolam (0.03 - 0.3 mg/kg) and bretazenil (0.003 - 0.1 mg/kg) dose-dependently increased rates of punished responding. Flumazenil (0.3 - 3.0 mg/kg) antagonized the rate-increasing effects of both bretazenil and midazolam. Other monkeys (N=3) responded under an FR30 schedule of stimulus shock termination. Midazolam decreased rate of responding (ED50z+0.3q0.1 mg/kg). Flumazenil (0.3-3.0 mg/kg) and bretazenil (0.01-1.0 mg/kg) surmountably antagonized the rate-decreasing effects of midazolam. Schild plot analysis revealed apparent pA2 values of flumazenil and bretazenil with midazolam which are in general agreement with reported affinity estimates of these drugs in vitro. Mean values (with 95% CL) were flumazenil with midazolam pA2 =7.09 (6.23, 7.95) with a slope of -1.04 (-1.49, -0.59); bretazenil with midazolam pA2 =7.77 (7.23, 8.310 WITH A SLOPE OF -0.94 (-1.06, -0.82). These results indicate that Schild analysis can be used to obtain quantitative measures of benzodiazepine effects in vivo. Moreover, our results demonstrate that under conditions in which bretazenil has no agonist effects it